LC
LINE CHECK LIVE Kitchen Operations OS
Explore Engine
STATION REPORT: Tested in Pendleton Catering Kitchens

Kitchen Chaos,
Resolved in Milliseconds.

A high-performance, offline-first inventory management application built on a modern AI-powered pipeline. Engineered to keep high-velocity catering environments synchronized, even in the heat of a 500-guest service.

9-Stage Invoice Engine
7-DB Parallel UPC Search
Offline IndexedDB Sync
linecheck-terminal-v1.0 LIVE
OPERATOR: CHEF AUSTIN
STATION: PENDLETON CATERING
SYS_STATUS: ACTIVE & PERSISTED
INVOICE PIPELINE STAGE 8/9
>_ assertInvoiceMath()
✔ sysco_invoice_1092.pdf passed
Total lines validated: 24
Math Guard: Sum verified ($1,482.50)
ACTIVE DEV INVENTORY 1,420 Items
High-Carbon Chef Knives 12 Qty
Panko Breadcrumbs 50 lbs
Premium Cue Tips 4 Packs

THE KITCHEN PRINCIPLE

"In a fast-paced kitchen, minutes are currency. I didn't build this app to write papers; I built it to keep our lines moving. When you are running a service, you cannot afford to wait on flaky network queries or manual database entries."
Austin Zumaya HEAD CHEF & LEAD ARCHITECT
PENDLETON CATERING
THE ORIGIN STORY

Made in the Fire.
Refined by Code.

Catering services are chaotic, dynamic, and unrelenting. As the Head Chef and Culinary Manager at Pendleton Catering in Oregon, Austin Zumaya knows this environment intimately. He is constantly on the move — coordinating high-volume prep sheets, checking temperatures, and managing vendor deliveries on iPad and Samsung mobile devices.

Traditional inventory software failed because it was built for desks, not prep tables. In response, Austin designed and engineered his own system. This is a technical solution tailored for gloved hands, LED glare, and dead network zones.

Built for Mobility: Native camera capture and fully responsive layouts optimized for iPad and Samsung Galaxy S25 Ultra devices.
Zero-Downtime Design: Complete offline queues utilizing IndexedDB to cache shelf inventories and invoice scans automatically.
TECHNICAL ARCHITECTURE

The Modern Production Stack

Every dependency of the Kitchen Operations OS was chosen for performance, type safety, and direct hardware integration in challenging kitchen conditions.

UI CORE

React 19 & TypeScript

Provides full type safety across all frontend components. High-efficiency UI renders ensure that inventory adjustments and scanned outputs display instantly.

AI CO-PILOT

Vertex AI & Gemini

Google's gemini-3-flash-preview powers invoice OCR, shelf analysis, and custom culinary calculations, wrapped in Vertex SDK secure proxies.

DATA PERSISTENCE

Firebase Cloud Platform

Firestore database with local offline persistence, secure Auth with strict kitchen manager timeouts, and scalable Storage for scanned invoice assets.

HARDWARE INTERFACE

Capacitor Native Bridge

Bridges web app compilation to native iOS and Android to hook directly into system-level cameras, custom storage directories, and physical scan behaviors.

STATE & STYLING

Zustand & Tailwind v4

Lightweight atomic global state management paired with high-performance utility-first styling designed for ultra-low responsive layouts.

OFFLINE ACCELERATION

IndexedDB & ZXing

Local IndexedDB queue stores offloaded scans and pushes auto-synced commits on reconnection. ZXing handles barcode decode logic.

WATERFALL PIPELINES

Rigorous 9-Stage Architectures

Both invoice extraction and barcode lookup systems utilize deterministic, multi-layered waterfall pipelines to guarantee absolute accuracy in rapid culinary workflows.

Invoice Processing Waterfall

Converts physical inventory invoices to verified digital assets without human math errors.

Primary Engine: gemini-3-flash-preview
ST_01
STAGE 1

Camera Capture

Chef captures a high-resolution photograph of the invoice using native device cameras (optimized for iPad/S25 Ultra).

ST_02
STAGE 2

Image Pre-processing

Converts image to grayscale, applies CLAHE for LED glare control, and triggers Sauvola Adaptive Binarization.

ST_03
STAGE 3

Vendor Detection

Performs matching on Tax ID, header texts, and signatures. Directs to: SYSCO, VERNS, US FOODS, or UNKNOWN.

ST_04
STAGE 4

Vendor-Specific Prompts

Applies targeted custom prompts. SYSCO logic handles sideways rotations. VERNS handles distinct colored sheets.

ST_05
STAGE 5

AI Extraction Core

Gemini parses image text. Automated OpenAI GPT-4o triggers instantly if Gemini is offline, complete with 3 retries.

ST_06
STAGE 6

JSON Normalization

Strips markdown syntax fences and structures extracted text into standard, uniform invoice schemas.

ST_07
STAGE 7

Structural Validation

Runs validateInvoice() validator logic. Throws errors and highlights incomplete fields immediately.

ST_08
STAGE 8

Math Guard Validation

Runs deterministic assertInvoiceMath(): validates qty × unit = total, and sums line items against total.

ST_09
STAGE 9

Manager Approval

Displays validated invoices on ActivityPage. Cloud Function pushes database items to inventory upon final approval.

DETAILED SPECIFICATIONS

Technical Specifications

For engineers and culinary professionals alike, here are the core parameters backing Line Check Live.

AI Engine & Security Setup

  • Primary LLM: gemini-3-flash-preview
  • Stable Model Option: gemini-2.5-flash
  • Fallback Model: OpenAI GPT-4o
  • Vertex AI integration: VertexAIBackend SDK proxy
  • Planned Sec update: Secret Manager Integration

Client State & Performance

  • State manager: Zustand (Atomic Singletons)
  • Local Cache Store: IndexedDB Key-Value Cache
  • Database: Firebase Firestore (offline active)
  • Image Processing: Sauvola Adaptive Binarization
  • Hardware Targets: iPad Air, Samsung Galaxy S25 Ultra

Deploying Speed to Your Production Kitchen

The landing page of the Pendleton Catering App, running on the registered production domain linecheck.store, represents a leap in professional kitchen management.

Registered at: IONOS Web Hosting Plus Active Site Directory: linecheck.store